function lst = freq_pat(H,pat)% freq_pat.m - generate frequency partition%% input:%   H:    half length of the frequency span, H needs to be greater than 16%   pat:  type of frequency partition which satsifies parabolic scaling relationship%         equal to 'p' or 'q'%% output:%   lst:  data representing the partition of frequency%% Written by Lexing Ying and Laurent Demanet, 2006    if(H<16)    error('H needs to be at least 16');  end    if(    pat=='p') %parabolic scaling by lexing    len = ceil(log2(H)/2)+1;    lst = cell(len,1);    lst{1} = [1 1];    lst{2} = [0 1];    lst{3} = [0 1 1 1];    cnt = 16;    idx = 3;    rad = 4;    while(cnt<H)      old = cnt;      lst{idx} = [lst{idx} 1 1];      cnt = cnt + 2*rad;      idx = idx+1;      rad = 2*rad;      trg = min(4*old,H);      lst{idx} = [zeros(1,cnt/rad), ones(1,(trg-cnt)/rad)];      cnt = trg;    end  elseif(pat=='q') %parabolic scaling by laurent    len = floor(log2(H)/2)+1;    lst = cell(len,1);    lst{1} = [1 1];    lst{2} = [0 1 1 1];    cnt = 8;    idx = 2;      rad = 2;    while(cnt<H)      old = cnt;      lst{idx} = [lst{idx} 1 1];      cnt = cnt + 2*rad;      idx = idx+1;      rad = 2*rad;      trg = min(4*old,H);      lst{idx} = [zeros(1,cnt/rad), ones(1,(trg-cnt)/rad)];      cnt = trg;    end  elseif(pat=='u')    len = ceil(log2(H)/2)+1; %uniform partitioning    lst = cell(len,1);    B = 2^(len-1);    lst{end} = ones(1,H/B);  else    error('wrong pat');  end